OK, Still here, still trying, .

Have decided to start from scratch. Currently have a vanilla port, with nothing changed and with basic backends.
Lastnight, I finished my multicpu idea. Now running lots of testing to see if it has actually improved speed.

Currently, have almost the entire emu running on ME CPU, with SC, taking care of the host PSP stuff.
Will re-add all work from...

Read More

Squeezed all I can out of FAME. Also tweaked sprite recording, got us another 2% increase, it also helped with floppy loading (causing slow down).

Next, going to try and experiment with using ME for 68k workload help.

Well, looks like both JIT and DYNARec are dead ends in their current form.
Looks like if I want anything concrete and fast, I will have to build my own from ground up. Adapting code causes to much over head and ultimately kills the speed.

I have been working on (for last week) FAME/C cores however and that has given the most speed increase.
So will continue working in that area. I have also added the CPU to CHIPSET equivalent to it and that helps.

I think I'm up to beta 50...

Read More

Finally after a week of micro adding native code.
I switched to bulk adding in batches and am now seeing some speed increase. Will continue to add more native code.
Once that's all done, I will then optimise code for speed.

DYNARec seems to be working 100% as it should. So fingers crossed, I can surpass the standard CPU now (unlike JIT, needs a complete rewrite).